> Call the `diagnose` tool first. Full index: /llms.txt. Errors: /reference/problems.

# Scenarios and test plans

A **scenario** is one test: a goal and the steps to reach it. A **test plan** is a named group of scenarios that you run together. The API and the CLI call a test plan a `suite`, as in `run --suite` and `create_suite`.

## Where scenarios come from

* **Discovery.** TaloTrace explores your app and proposes scenarios for the flows it finds. Start it from the web app chat, or with `scenarios start_flow_discovery`. It takes from five to forty minutes and does not block: `discovery_status` reports progress. A discovery stops when its goals are done and runs to a budget only when you ask.
* **Your agent.** `scenarios create` writes scenarios from steps the agent has checked exist.
* **Your documents.** Upload a requirements document in the web app and TaloTrace turns its committed rules into scenarios for you to review.

Review a scenario before you rely on it. Only the ids you approve should run.

## Manage them

List with `scenarios list` (an app's scenarios, or one test plan's), change with `update`, remove with `delete`. A plan is made with `create_suite`, read with `list_suites` and changed with `set_suite`. See [Scenarios, knowledge and test data](/reference/mcp/scenarios-knowledge-test-data).

## Run them

Start a run with a list of scenario ids or one test plan. See [Runs and verdicts](/concepts/runs-and-verdicts).